HP-UX Virtual Partitions

By Marty Poniatowski

An Overview of IEEE802.3, TCP/IP

In order to understand how the networking on your UNIX system works, you first need to understand the components of your network that exist on your UNIX system. Seven layers of network functionality exist on your UNIX system, as shown in Figure 13-2. I cover the bottom four layers at a cursory level so that you can see how each plays a part in the operation of your network and, therefore, be more informed when you configure and troubleshoot networking on your UNIX system. The top layers are the ones that most UNIX system administrators spend time working with because those layers are closest to the functionality to which you can relate. The bottom layers are, however, also important to understand at some level, so that you can perform any configuration necessary to improve the network performance of your system, which has a major impact on the overall performance of your system.

I start reviewing Figure 13-2 at the bottom with layer 1 and then describe each of the four bottom layers. This is the International Standards Organization Open Systems Interconnection (ISO/OSI) model. It is helpful to visualize the way in which networking layers interact.

Physical Layer

The beginning is the physical interconnection between the systems on your network. Without the physical layer, you can't communicate between systems, and all the great functionality you would like to implement is not possible. The physical layer converts the data you would like to transmit to the analog signals that travel along the wire (I'll assume for now that whatever physical layer you have in place uses wires). The information traveling into a network interface is taken off the wire and prepared for use by the next layer.

Link Layer

In order to connect to other systems local to your system, you use the link layer that is able to establish a connection to all the other systems on your local segment. This is the layer where you have either IEEE 802.3 or Ethernet. Your UNIX system supports both of these "encapsulation" methods. This is called encapsulation because your data is put in one of these two forms (either IEEE 802.3 or Ethernet). Data is transferred at the link layer in frames (just another name for data), with the source and destination addresses and some other information attached. You might think that because two different encapsulation methods exist, they must be very different. This assumption, however, is not the case. IEEE 802.3 and Ethernet are nearly identical. For this reason, many UNIX systems can handle both types of encapsulation. So with the bottom two layers, you have a physical connection between your systems and data that is encapsulated into one of two formats with a source and destination address attached. Figure 13-3 lists the components of an Ethernet encapsulation and makes comments about IEEE802.3 encapsulation where appropriate:

One interesting item to note is the difference in the maximum data size between IEEE 802.3 and Ethernet of 1492 and 1500 bytes, respectively. This is the Maximum Transfer Unit (MTU). The ifconfig command covered shortly displays the MTU for your interface. The data in Ethernet is called a frame (the re-encapsulation of data at the next layer up is called a datagram in IP, and encapsulation at two levels up is called a packet for TCP).

Keep in mind that Ethernet and IEEE 802.3 will run on the same physical connection, but there are indeed differences between the two encapsulation methods. With your UNIX systems, you don't have to spend much, if any, time setting up your network interface for encapsulation.

Network Layer

Next we work up to the third layer, which is the network layer. This layer on UNIX systems is synonymous with the Internet Protocol (IP). Data at this layer is transported as datagrams. This is the layer that handles the routing of data around the network. Data that gets routed with IP sometimes encounters an error of some type, which is reported back to the source system with an Internet Control Message Protocol (ICMP) message. We will see some ICMP messages shortly. ifconfig and netstat are two UNIX commands that are commonly used to configure this routing.

Unfortunately, the information that IP uses does not conveniently fit inside an Ethernet frame, so you end up with fragmented data. This is really re-encapsulation of the data, so you end up with a lot of inefficiency as you work your way up the layers.

IP handles routing in a simple fashion. If data is sent to a destination connected directly to your system, then the data is sent directly to that system. If, on the other hand, the destination is not connected directly to your system, the data is sent to the default router. The default router then has the responsibility of getting the data to its destination. This routing can be a little tricky to understand, so I'll cover it in detail shortly.

Transport Layer

The trasport level is the next level up from the network layer. It communicates with ports. TCP is the most common protocol found at this level, and it forms packets that are sent from port to port. The port used by a program is usually defined in /etc/services, along with the protocol (such as TCP). These ports are used by network programs such as telnet, rlogin, ftp, and so on. You can see that these programs, associated with ports, are the highest level we have covered while analyzing the layer diagram.